Meybe someday GPUs(and all)will use this-Replace Transistors
- Thread starter pc999
- Start date
Killer-Kris
Regular
reever said:
I don't remember where, but I was hearing hz speeds with the current experiments, of course that should definitely be expected to improve with time.
I think the big deal with these is that they can be made so incredibly small. Currently far beyond anything imaginable for transistors.
Mate Kovacs
Newcomer
One Cell to rule them all. 
I was reading up on this. If you go to hp's website, they link to the academic papers that describe what this press release is actually about (but I think you'd have to either pay for it or be at a University to get access). It's kind of interesting as it's a switching system which uses atoms-thick devices using no silicon that could be made into logical units. But this was basically a feasibility test at this stage, as I didn't see the paper getting into switching speeds or maximal densities, though it looked like the current densities of these devices were much lower than those for transistors. Also an issue, in my mind, was that the device worked between about +1.5V and -1.5V, which seems like it'd require a fair amount of energy to switch, and thus not be terribly fast. Anyway, I'm sure I could check it better if I was on campus atm, but I can't view the pdf from home (I didn't save it to my computer there).
Edit: On campus now, so I gave it another glance. The main pull of this sort of thing is that with enough engineering work, this sort of thing could eventually end up smaller/faster than transistors. But the current implementation is quite a bit larger, and vastly slower (looks like they tested the circuit in frequencies of 3-4 Hz). So it may be a while, and it may be possible that this implementation won't be useful at all. But, the paper's authors are presenting the crossbar latch as a general computing tool that could be used with a variety of different types of core.
On a personal note, though, I think that half-metals really seem like they'll be the useful material for moving computers past semiconductors.
Edit: On campus now, so I gave it another glance. The main pull of this sort of thing is that with enough engineering work, this sort of thing could eventually end up smaller/faster than transistors. But the current implementation is quite a bit larger, and vastly slower (looks like they tested the circuit in frequencies of 3-4 Hz). So it may be a while, and it may be possible that this implementation won't be useful at all. But, the paper's authors are presenting the crossbar latch as a general computing tool that could be used with a variety of different types of core.
On a personal note, though, I think that half-metals really seem like they'll be the useful material for moving computers past semiconductors.
rendezvous
Regular
Mintmaster
Veteran
Without knowing the current used, how can you say that?Chalnoth said:
For example, in electrochemistry, you need the same voltage to switch no matter what the concentration or volume of electrolyte. But you usually have over 10^20 molecules that are switching. If you could make a switch that only needs a handful of electrons, power consumption could be a non factor. The density of heat generation is another issue, but switching voltage is not really indicative of anything other than potential leakage current.
Having researched molecular electronics myself, I can say that a bigger problem is reliability & reproducability. You need switches that will work over 10^15 times (just a guess) without failing.
Voltage difference relates directly to the energy difference between two states, though. A large voltage difference means it'll take a large amount of energy to switch the device, which has nothing to do with leakage current (i.e. in an idealized system, the majority of the energy consumption will be in switching, which is rather unlike today's semiconductor devices)Mintmaster said:
But, to tell you the truth, the more telling number is the switching speed that they used for the experiment: ~3Hz. That said, there's no reason why the particular materials that hp used for this experiment need be used in future devices.
rendezvous said:
